P h a s e th eor y an d p ro s odi c s pe llou t: T h e c...

Phase theory and prosodic spellout:The case of verbs1

ANGELIKA KRATZER AND ELISABETH SELKIRK

The Linguistic Review 24 (2007), 93–135 0167–6318/07/024-093DOI 10.1515/TLR.2007.005 c©Walter de Gruyter

Abstract

In this article we will explore the consequences of adopting recent proposalsby Chomsky, according to which the syntactic derivation proceeds in terms ofphases. The notion of phase – through the associated notion of spellout – allowsfor an insightful theory of the fact that syntactic constituents receive defaultphrase stress not across the board, but as a function of yet-to-be-explicatedconditions on their syntactic context. We will see that the phonological evi-dence requires us to modify somewhat the theory of which functional categoriesactually define a phase. Patterns of default, syntax-determined, phrase stressare argued to result from prosodic spellout requiring the highest phrase in thespellout domain to correspond to a major prosodic phrase in phonological rep-resentation, and carry major phrase stress.

1. Research for this article was supported in part by NSF Grant #BCS-0004038 to ElisabethSelkirk. We would like to thank the many people who have given us feedback on this worksince we first began to develop it in a joint UMass seminar on meaning and intonation inSpring 2004 and later presented it at the Mediterranean Syntax Meeting June 2005, the Work-shop on Prosodic Phrasing at GLOW 2006 in Barcelona, the Prosody-Syntax Interface Work-shop, University College London, October 6, 2006, NELS 2006 at the University of IllinoisChampaign Urbana October 2006, and in Linguistics Colloquia at Tokyo University in March2006 and at the University of Potsdam in January 2007. In particular, we’d like to thank ShinIshihara, Caroline Féry, Kyle Johnson, Arsalan Kahnemuyipour, Carlos Gussenhoven, HubertTruckenbrodt, Sónia Frota, Michael Wagner, Gorka Elordieta, Vieri Samek-Lodovici, DavidAdger, Molly Diesing, Pilar Prieto, Jon Nissenbaum, Steve Franks, Karlos Arregi, Ingo Feld-hausen, Laura Downing, Anthi Revithiadou, Mari Takahashi, Malte Zimmermann, GisbertFanselow and Bob Franks, and an anonymous reviewer.

94 Angelika Kratzer and Elisabeth Selkirk

1. Introduction: Classic puzzles in the pitch accenting of verbs

There are a number of classic puzzles concerning the circumstances underwhich verbs may or may not bear neutral, default phrase stress in languageslike English, German or Dutch. Earlier approaches characterized these puz-zles in terms of the ability of a verb to bear an intonational pitch accent (e.g.,Schmerling 1976; Gussenhoven 1983, 1992; Selkirk 1984, 1995), but more re-cent work (e.g., Ladd 1996; Féry and Samek-Lodovici 2006; Selkirk 2006a)suggests that the presence of pitch accents reflects patterns of phrase stress. Soit is the variation in the ability of verbs to bear default phrase stress in neutralsentences that needs to be explained. We will argue that verbs in certain syn-tactic configurations may simply fail to be organized into the constituents overwhich phrase stress is defined in neutral, all-new, sentences while in other con-figurations they may themselves constitute such a constituent (Kahnemuyipour2004; Selkirk and Kratzer 2005; Selkirk 2006c). The observed organization ofverbs into phrase stress domains has a ready account in terms of phase-basedspellout.

A central question is why there should be a difference between verbs andtheir arguments in their potential for pitch accenting or phrase stress. In an all-new neutral sentence where an argument of the verb is present, the argumentDPs will necessarily show a pitch accent. But the pitch accenting/stressing pos-sibilities of the verb are more varied. One puzzling fact involves verbs that aretransitive, which in all-new sentences never necessarily carry an accent. As(1a) shows, a pitch accent may be present on a transitive verb in medial po-sition in an all-new sentence, though its presence is only optional there (pitchaccents are marked by ‘´’ over the vowel). This is true both in German and inEnglish. When the verb is in final position, which is the required order in Ger-man embedded sentences, there is no pitch accent on the verb in neutral all-newsentences, as in (1b). A pitch accent on studiert in (1b) would necessarily leadto a contrastive interpretation.

(1) Pitch accent in all-new, neutral sentences with transitive verbsa. Transitive verbs in medial position: optional pitch accent

MaríaMaria

studiert/studíertis studying

diethe

Gesétze.laws

‘Mária is studying/stúdying the láws.’b. Transitive verbs in final position: no pitch accent (if not con-

trastive)IchI

glaube,think

dassthat

MaríaMaria

diethe

Gesétzelaws

studiert/#studíert.is studying

‘I think that Maria is studying the laws.’

Phase theory and prosodic spellout 95

c. Verbal arguments in any position: necessary pitch accent(see (1ab))

Facts such as these have been widely observed and have been given widelyvarying accounts (see, e.g., Gussenhoven 1983, 1992; Selkirk 1984, 1995;Cinque 1993; Truckenbrodt 1995; Wagner 2005; Féry and Samek-Lodivici2006). This three-way distinction in possibilities of pitch accenting in neutralsentences is a reflection, we now think, of a three-way distinction in phrasestress. The appearance of pitch accent would be associated with the appear-ance of some degree of phrase stress (Ladd 1996; Féry and Samek-Lodivici2006; Selkirk 2006a). Where pitch accent is categorically required, as withverbal arguments in (1ab), a major phrase stress would be present. Cases ofoptional pitch accent, as in (1a), would show the optional presence of a minorphrase stress. Where pitch accent is categorically excluded, as in (1b), the rele-vant word would bear neither major nor minor phrase stress. For the transitiveverbs, then, the problem would be to explain, first, the categorical absence ofmajor phrase stress/necessary pitch accent and, second, the order-based alterna-tion seen in (1ab) between the presence and absence of optional minor phrasestress/pitch accent. We will argue in Section 3 that the distribution of majorphrase stress/necessary pitch accent in all-new sentences is determined by prin-ciples of phase-based spellout (namely, by principles of the syntax-phonologyinterface). The distribution of minor phrase stress, on the other hand, is appar-ently a matter for the phonology per se, and, specifically, principles of prosodicstructure organization (cf. Section 4).

Already, the well-known fact that certain, but only certain, intransitive verbsin sentence-final position must bear a pitch accent in all-new neutral sentencesindicates that verbs as a class do not have a uniform prosodic treatment, andthat syntactic and/or semantic properties play a role in determining their pros-ody.

(2) Pitch accent in all-new sentences with intransitive verbs/predicatesa. Eventives: no pitch accent on verb

(i) IchI

hab’have

geradejust

imin.the

Radioradio

gehört,heard

dassthat

derthe

Königking

vonof

BáyernBavaria

ertrunkendrowned

ist.is

‘I just heard on the radio that the King of Bavária hasdrowned.’

(ii) IchI

hab’have

gelesen,read

dassthat

diethe

Metállarbeitermetal

gestreiktworkers

haben.gone.on.strike have

‘I read that the metal workers went on strike.’


b. Statives: pitch accent on verb(i) Ich

Ihab’have

irgendwosomewhere

gelesen,read

dassthat

derthe

Königking

vonof

BáyernBavaria

gespónnenwas.crazy

hat.

‘ I read somewhere that the King of Bavária was crázy.’(ii) Ich

Ihab’have

gehört,heard

dassthat

derthe

RhéinRhine

stínkt.stinks

‘I’ve heard that the Rhíne stínks.’

In the sentences in (2b), a pitch accent is required on the verb in an all-newneutral sentence. Absence of a pitch accent on the verb in these cases wouldonly be possible if the verb itself were old information, given in the discourse:

(3) a. Ich hab’ irgendwo gelesen, dass der König von Báyern gespon-nen hat.(gesponnnen must be given information)

b. Ich hab’ gehört, dass der Rhéin stinkt.(stinkt must be given information)

In her treatment of the prosodic contrasts in comparable intransitive sentences,Diesing (1990) argues that the presence/absence of pitch accents on intransitiveverbs in all-new neutral sentences is a function of the position of the subjectof the verbs in syntactic structure, which at least in the case of unergatives, de-pends on whether the verb is an eventive (“stage-level”) or stative (“individual-level”) predicate. If this sort of analysis of different subject positions for differ-ent types of intransitives is correct, it supports the idea that obligatory phrasestress is assigned (or not assigned) to a verb in function of hierarchical syntac-tic structure, and not merely in terms of its position in the word order of thesentence.

Section 3 will make the case that the aspects of syntactic structure whichare relevant to defining default patterns of necessary pitch accent/phrase stressfind an insightful characterization in terms of the theory of phases and multiplespellout. Section 4 addresses the role for phrase stress in characterizing thedistribution of pitch accents and the role for prosodic phrasing in the generationof default phrase stress.


2. Background

2.1. Focus-related phrase stress

There are two semantic/pragmatic properties that have an impact on phrasestress. One is the given/new organization of the sentence, which is arguablyrepresented via an interpretable G-feature on given constituents (cf. Féry andSamek-Lodovici 2006; Selkirk 2006a, b). The other is contrastive focus, whichwe take to be represented via an interpretable F-feature (cf. Jackendoff 1972;Rooth 1992, 1996) that may be borne by a constituent in syntactic structure. Inthis article, we are interested in principles of stress assignment that are inde-pendent of G-marking or F-marking, and this is why we are primarily lookingat syntactic representations that do not contain given or contrastively focusedconstituents, hence are “neutral”. Ultimately, the interaction between princi-ples spelling out the syntactic-constituent-structure-dependent “default” phrasestress and F- and G-marking-dependent principles of phrase stress has to betaken into account. The full array of stress patterns we observe in language isthe result of this interaction.

It is a well-known fact about English, German and Dutch that a discourse-given phrase may fail to receive a pitch accent. This is illustrated in the dis-course in (4):

(4) A: Ánscombe has been féuding with her cólleagues.B: Wíttgenstein brought a glass of whískey over to AnscombeG.

Perháps they have made úp.

In the B response to A, there is no pitch accent (or phrase stress) on Anscombe,which has been used in the previous sentence in the discourse. But there isa necessary pitch accent (and phrase stress) on Wittgenstein and (a glass of)whiskey, which are new in the discourse.

The necessary absence of pitch accent/phrase stress on a discourse-givenentity is argued in recent work by Féry and Samek-Lodovici (2006) and Selkirk(2006a, b) to be the consequence of a G-marking in the syntax for discourse-given entities and an interface principle Destress Given, part of spellout, whichcalls for the absence of phrase stress on G-marked elements:

(5) Destress Given (Féry and Samek-Lodovici 2006; see also Selkirk2006a)A given phrase is prosodically non-prominent.

Destress Given plays a role in an additional puzzle about the pitch accent-ing/stressing of verbs. Consider the discourse in (6), which is a variant of aclassic example of Ladd’s (1980, 1996):


(6) A. Why don’t you háve [some Frénch tóast]?B. I’ve forgótten how to máke [French toast]G.B′. (#I’ve forgótten how to make [French toast]G.)

The verb bears a pitch accent/phrase stress when the object DP is discourse-given, as in (6B), but we see in (6C) that, if the verb make also lacks a pitchaccent when the direct object is given, the verb and the entire verb phrase wouldbe interpreted, erroneously, as given in the discourse. This means that when thedirect object lacks phrase stress, the adjacent transitive verb here necessarilybears phrase stress if it is discourse-new. Compare this to the case where asimilarly discourse-new transitive verb like have appears before a discourse-new and accented/phrase stressed direct object, shown in (6A). In this case, theappearance of a pitch accent on the discourse-new verb is optional, not neces-sary. There are similar facts in German, where the so-called deaccenting of adiscourse-given direct object can also entail the presence of pitch accent/phrasestress on the transitive verb. This is true regardless of the relative order of theverb and the direct object, as the B/B’ sentences in the discourse in (7) show:

(7) Verb with G-marked direct objectA. I have great respect for the laws.B. María

Mariastudíertis studying

diethe

GesetzeG.laws

‘Mária is stúdying the laws.’B′. Ich

Ihab’have

gehört,heard

dassthat

MaríaMaria

diethe

GesetzeGlaws

studíert.is studying

‘I’ve heard that Mária is stúdying the laws.’

We assume that the simple prominence spellout principle Destress Given is re-sponsible for the absence of phrase stress on the discourse-given DPs in thesentences above. But we will have to explain why main phrase stress is there-fore required to appear on the verb adjacent to the G-marked direct object. Thiswill be taken up in Section 3.

The second principle of focus-related prominence spellout involves con-trastive focus. The term “contrastive focus” will be used here to designate thestatus of a constituent in sentences like I gave one to SARAH, not to CAITLIN,or I only gave one to SARAH, where the meaning of the sentence includes aspecification that there exist alternatives to the proposition expressed by thesentence which are identical to that proposition except for different substitu-tions for the contrastively focused constituent.2 The alternatives set here would

2. This type of focus is referred to variously as contrastive focus, identificational focus, alterna-tives focus, or simply focus (Jackendoff 1972; Jacobs 1988; Krifka 1992; Rooth 1992, 1996;Kiss 1998; Kratzer 2004).


include {I gave one to Sarah, I gave one to Caitlin, I gave one to Stella, . . . }.This type of focus has a direct role in determining the semantic interpretationof the sentence, affecting truth conditions and conversational implicatures. Italso has an effect on stress patterns. Truckenbrodt (1995) and Rooth (1996)independently proposed a principle for the phonological interpretation of con-trastive focus, which we consider here to be a principle of prosodic spellout.

(8) Contrastive Focus Prominence Rule (CFPR; Truckenbrodt 1995;Rooth 1996)Within the scope of a focus interpretation operator, the correspondingF-marked [contrastive focus] constituent is the most metrically promi-nent.

Subsequent work by Féry and Samek-Lodovici (2006) and Selkirk (2006a,b)has shown this principle to have very desirable consequences, and we adopt ithere. Its effect on verbs is straightforward. Like virtually any other constituentin a sentence of English, a verb that is F-marked can bear the stress prominencecalled for by the Contrastive Focus Prominence Rule (CFPR). If that F-markedverb appears in a sentence containing other discourse-new constituents thatbear a phrase stress, as in (9), the CFPR predicts that the F-marked verb wouldcarry a yet higher level of stress, namely sentential stress, which would make itmost prominent within the domain. (9) contains a case of contrastively focusedverbs in a right-node raising construction in English (Selkirk 2002; underlineindicates the greater stress that is associated with contrastive focus, cf. Katzand Selkirk 2006):

(9) A. What evidence do you have that the mother of the bride wasstressed out?

B. She assémbledF, and then lóstF the invitátions.

In this case, default stress assignment is responsible for the normal phrase stressthat appears on the discourse-new direct object, while prominence spellout, inthe form of the CFPR, requires the additional presence of phrase stress at aneven higher level on the verb. There is no optionality to the phrase stress/pitchaccent on the verb in (9). Compare the case where lost is not a contrastivefocus:

(10) A. What’s going on?B. My móther lost/lóst the invitátions!

Here, as in (1a), the pitch accent on the medial transitive verb is optional, re-flecting an absence of major phrase stress.

In sum, the theory of prosodic spellout needs to include the G-marking andF-marking spellout principles Destress Given and Contrastive Focus Promi-


nence Rule, which give rise to stress patterns that are not predicted by the de-fault principles of phrase stress on their own.3 For this reason, in our analysisof default phrase stress patterns below, we will largely confine our attention tosentences that are “neutral” in that they lack any contrastive focus or discourse-given constituents.

2.2. Prosodic structure representation of phrase stress

Neither Destress Given nor the CFPR presupposes any particular theory ofthe phonological representation of phrase stress. Whatever phrase stress is,the CFPR says that an F-marked element must have a higher level of it thanother elements in the same focus scope. And whatever phrase stress is, De-stress Given says that a G-marked element can’t have any. However, the inves-tigation of the interaction of CFPR and Destress Given with the principles ofdefault phrase stress by Féry and Samek-Lodovici (2006) and Selkirk (2006a)provides evidence that we have to distinguish levels of phrase stress that are de-fined in terms of phrases of the prosodic hierarchy, namely intonational phrase,major phrase and minor phrase, cf. Section 4. So in what follows, we opt tounderstand the patterns of default phrase stress in terms of that prosodic phras-ing.

Evidence for a theory of phonological representation which includes a hi-erarchical prosodic phrasing structure that is independent of but related to thesyntactic structure of the sentence is provided by a wide range of segmen-tal and suprasegmental phenomena, both phonological and phonetic (see, forexample, Nespor and Vogel 1986; Selkirk 1986, 2003; Inkelas and Zec 1990;Truckenbrodt 1999; Frota 2000; as well as the articles in the current issue). Ac-cording to the prosodic structure theory of stress (Selkirk 1980; Hayes 1981,1995; Nespor and Vogel 1986, and other more recent work), the different typesof prosodic constituents into which a sentence is parsed in phonological repre-sentation may each be prosodically headed, and the notion “stress” is definedin terms of this prosodic headedness. A “stressed syllable” is the syllable thatis the head of a foot, for example, while a syllable that carries “word stress” or“main word stress” is the head of the foot that is the head of the prosodic word.This sort of stress representation can be depicted with a prosodic-constituent-bracketed metrical grid (Halle and Vergnaud 1987; Hayes 1995), in which theposition of a grid mark x within a constituent marks the locus of the head-prominent, stressed constituent at the next level down in the prosodic structure:

3. For the interaction of these principles when a constituent is both F-marked and G-marked, seeSelkirk (2006a).


(11) ((bxal.l

)e(.r

xxi.n

))a =

prosodic wordfootballerina

“(main) word stress”“stress”

Above the level of the word, prosodic-structure theorists (Selkirk 1978, 1986,2005; Nespor and Vogel 1983, 1986; Beckman and Pierrehumbert 1986, Pier-rehumbert and Beckman 1988, and others) concur in isolating at least two dis-tinct levels of prosodic phrasing which play a role in the phonology and whichhave a reliable relation to syntactic constituency – the intonational phrase anda lower level of prosodic phrasing variously referred to as phonological phrase,major phrase, or intermediate phrase.4 We’ll use the term “major phrase” forthis level and “major phrase stress” to refer to the head of major phrase (MaP).“Sentential stress” and “intonational phrase stress” can be used interchange-ably to refer to the higher level of phrase stress found in intonational phrase(IP).

(12) (((xxófficers

)) (

escxorte

)d

((baller

xxxí na

)))sF

intonational phrase (“sentential stress”)

major phrase (“major phrase stress”)

prosodic word

The syntax-driven default theory of phrase stress with which this article is con-cerned can arguably be understood as a theory of phrase stress at the majorphrase level. In the materials investigated for English, each of the stressesaccompanied by a necessary pitch accent in English is separated by a majorphrase (= intermediate phrase) boundary (Selkirk 2006a), whose mark in En-glish declaratives is the presence of a peripheral low (L-) phrase edge tone(Beckman and Pierrehumbert 1986).

Within the prosodic phrases of phonological representation, according to thistheory, properly phonological constraints determine the presence and positionof the prosodic heads which represent the main stress of those phrases (cf.Section 4). Patterned on the theory of word stress, the prosodic theory of phrasestress leaves it as a language-particular option whether the main stress withina major phrase will fall at the right or left edge of that phrase. In English andGerman, for example, the main stress of major phrase falls on the right. Amajor phrase consisting of an adjective followed by a noun will carry mainstress on the noun:

4. An additional lower level of phrasing, variously referred to as minor phrase or accentualphrase, has also been isolated in Japanese (Poser 1984; Kubozono 1993) and Korean (Jun1995, 1998) and will play a role in Section 4 in our analysis of pitch accenting in English andGerman.


(13) English and German: The head of major phrase is on the righta.

[

(

The yoxung baller

xxina

)

s] hxate [

(

the nxew l

xxaw

)

s]

major phrase

prosodic word

b.

[

(

Die jxungen Baller

xxina

)

s] hxassen [

(

die nxeuen Ges

xxetz

)

e]

major phrase

prosodic word

In Persian, by contrast, default main stress of the major phrase falls on theleftmost element (Kahnemuyipour 2003):

(14) Persian: The head of major phrase is on the left(

[

xxin d

xo ket

xâb

)

]

major phrase

prosodic word

‘these two books’

With this theory, then, a prosodic phrasing organization would be establishedas part of spellout. Within prosodic constituents, the words must have beenlinearized into the order that appears on the surface, a linearization which pre-sumably also is part of spellout. And any main stress/prosodic head within theprosodic constituent is determined by the phonological principles for definingthe presence and rightmost or leftmost position of that main stress. Accord-ing to this theory of default stress assignment, then, any influence of syntacticstructure on the location of default stress in neutral sentences is understood asa matter of how the prosodic phrasing is spelled out on the basis of the syn-tax, not as a direct assignment of stress on some syntactically defined domain(cf. Section 3.1 below). It is also predicted that the same spelled-out prosodicphrasing that is appropriate for characterizing the distribution of phrase stressin the sentence provides the domains over which other phonological and pho-netic phenomena are defined as well, whether tonal or segmental. (See, e.g.,Hayes and Lahiri 1991 for evidence of this sort from Bengali.)

Earlier theories of the relation between syntax and prosodic structure haveunderstood that relation to be established in non-derivational fashion, and en-tirely crosscategorially, on the final output of the syntactic component (e.g.,Selkirk 1986, 1995, 2002, 2003, 2005; Nespor and Vogel 1986; Truckendbrodt1999). But recent developments in syntax, in particular the theory of phases,encourage the examination of the phonological effects of (i) derivationally in-terweaving the syntactic derivation and the interface with phonology, and (ii)privileging only certain syntactic configurations, namely the syntactic comple-ments of phase heads, as the locus of that spellout. The phenomena that we willexamine below certainly do argue that the syntax-prosodic structure interface is


not blandly cross-categorial but involves relations between prosodic structureand only certain constituents of the extended verbal projection, as predicted byphase theory.

2.3. Phases and spellout: The basics

Phases provide the infrastructure of a theory of the syntactic derivation. ForChomsky (2000, 2001, 2005a, b), the phrase νP, headed by the external argu-ment-introducing little ν, and the phrase CP, headed by the complementizerC, correspond to phases. Within a phase, lexical material is inserted and con-stituents may move up to higher phase-internal syntactic positions. At the endof a phase, the material in the complement to the head of the phase is spelledout. It is during spellout that phonological form is given to words. Accordingto recent proposals, including the present one, it is also during spellout that anyhigher order prosodic structure, be it phrase stress or prosodic phrasing, wouldbe assigned. Because there may be more than one phase in the derivation ofthe sentence, there may be more than one instance of spellout, hence the termmultiple spellout for the phasal derivation of phonological form.

Figure 1 illustrates the theory of phases and spellout as it applies to (15).

(15) Maria studied laws.

The phrases corresponding to phases are indicated with the solid curved lines.Spellout will assign a phonological representation to the elements of the spell-out domain in each phase, indicated by the dotted curved line. The spelloutdomain of a phase is the complement of the phase head. The lowest spelloutdomain in Figure 1 is then the VP complement of the lowest phase head, whichis ν. Spellout takes place during the νP phase on the VP, which contains thedirect object laws and the verb. The subject of a transitive sentence, which isgenerally assumed to be introduced in the Spec of νP, is unaffected by spelloutduring the νP phase; it is not in the VP spellout domain of that phase. On thenext higher phase in Figure 1, which is the CP phase, the not-yet-spelled-outsubject might rise to Spec of TP and will be assigned a phonological represen-tation at that point, since TP, as the complement of C, is the spellout domainof the CP phase. Subject and object of transitive sentences like (15) are thusspelled out in different phases, and this has consequences for the prosody, aswe will see.

The notion of phase provides a new way of thinking about the cycle in bothsyntax and phonology. For phonology what stands out is the idea that not allsyntactic constituents are “cyclic domains” for phonological realization, ratherit is only those that are defined in terms of phase-based spellout. Our purpose inthis article is to understand the manner in which prosodic phrasing and conse-


quent default phrase stress are assigned to a sentence. Phase theory leads to thehypothesis that such properties will be produced as a consequence of prosodicspellout on phase-dependent spellout domains.

3. A phase-based theory of default phrase stress

3.1. Transitive sentences: What constituents are prosodically spelled out?

The simplest, minimal, theory of prosodic spellout would simply say that thespellout domain of a phase, namely the complement of a phase head, corre-sponds to the domain over which phrase stress is defined.

(16) Prosodic spellout: The simplest theoryThe spellout domain of a phase is the prosodic domain for phrasestress.

CPCP phase

SpecCP

C TP

Spellout on CP phase

SpecTP

T νP

νP phase

Spec νP

Maria

ν VP

Spellout on νP phase

V

studied

DP

laws

Figure 1. An illustration of phases and spellout for Maria studied laws.


Adger (2006) has made this proposal, assuming that phrase stress is assigneddirectly on the syntactic structure. Ishihara (2007) has proposed that the spell-out domain of a phase corresponds to a prosodic major phrase in phonolog-ical representation, in order to account for certain non-stress phenomena ofJapanese sentence tonology. But Kahnemuyipour (2004) offers an importantcritique of the simple theory in (16). It cannot account for the fact that, univer-sally, in the default stress pattern of all-new transitive sentences, phrase stressalways falls on the direct object, never the verb, regardless of the order in whichthey appear:

(17) Kahnemuyipour’s generalizationNo phrase stress on a transitive verb in all-new sentences.

Universally: * [ object vxerb ] and * [v

xerb object]

As Kahnemuyipour points out, this fact is not predicted by a theory that placesthe verb and the direct object in the same domain for assignment of phrasestress, since, within that domain, prosodic theory would allow for either right-most or leftmost placement of main phrase stress (as seen with (13) vs. (14)).

Kahnemuyipour (2004) proposes that it is not the entire spellout domain ofthe phase which forms the domain on which phrase stress is assigned, but ratherthat only the highest constituent within the spellout domain is the domain withrespect to which phrase stress is assigned. But this characterization does notpick out the direct object as the highest constituent within VP in the configu-ration of Figure 1, for example, unless the object moves into a higher positionwithin the VP itself, which might not always be the case. Kahnemuyipour’scondition would also not pick out the PP as the highest constituent in caseswhere a VP consists of just a PP and the verb. It seems that a slightly differentformulation of Kahnemuyipour’s condition is needed. We suggest that the gen-eralization is instead that only the highest phrase within the spellout domain isselected for prosodic spellout. Direct objects and VP-internal PPs would there-fore be eligible for prosodic spellout but verbs would not, since they are notphrasal. We therefore arrive at the formulation in (18).

(18) The Highest Phrase Condition on prosodic spellout (cf. Kahnemuyi-pour 2004; also Selkirk and Kratzer 2005) – stress-based versionAssign phrase stress within the highest phrase within the spellout do-main.

Example (19) provides an example of how this proposal would work:


(19) [ . . . [ ν [xobject verb

︸︷︷︸νP phase: spellout domain

]VP ] ]νP →spellout

xobject verb

VP is the spellout domain of the lowest phase νP. Within that spellout domain,according to (18), only the highest phrase in the spellout domain will receivephrase stress. The direct object is the highest, and in this case, the only, phrasewithin VP.

We would like to suggest a further friendly revision to the Kahnemuyipour(2004) proposal, one which states prosodic spellout in terms of prosodic phras-ing, not phrase stress. The proposal is that the highest phrase within the spelloutdomain is spelled out as a prosodic major phrase.

(20) The Highest Phrase Condition on prosodic spellout – phrasing-basedversionThe highest phrase within the spellout domain of a phase correspondsto a prosodic major phrase in phonological representation.

In proposing (20), we assume that phonological principles that account for thepresence and location of main stress (the prosodic head) within a prosodic con-stituent provide the phrase stress patterns at issue. The phrase stress facts to bereviewed in Section 3 do not choose between the phrasing-based condition (20)and the stress-based condition (18). But it can be argued that further theoreticaland empirical considerations concerning the phonology favor phrasing-basedprosodic spellout. (See Section 2.2 and Section 4.) The next paragraphs reviewthe positive predictions of the Highest Phrase Condition on prosodic spelloutin terms of the phrasing-based formulation, first looking at what is predictedwithin the spellout domain of the lowest phase in the sentence, namely, withinthe VP that is complement to the phase-head ν (underlining marks the highestphrase within the spellout domain).

(21) Direct object, not the verb, gets phrase stress in an all-new sentence

[ . . . [ [ [xobject verb


]VP ν ] ]νP →spellout

(xobject )MaP verb

Spellout on the VP, which is the spellout domain on the ν-phase, will parse thedirect object as a major phrase (inside of which the phonology will assign mainphrase stress). The verb will be spelled out segmentally, but will not be orga-nized into a prosodic phrase. The Highest Phrase Condition correctly predicts


that the verb will fail to get phrase stress, while the direct object will indeedget stressed.

The Highest Phrase Condition on prosodic spellout also predicts the patternof phrase stress observed with VP-internal prepositional phrases in German(Uhmann 1991; Jacobs 1993). When a directional or locative PP intervenesbetween the direct object and a verb in final position, it does not get phrasestress, as seen in the cases in (22), but when the VP does not contain a directobject, a low PP will receive the phrase stress, as in (23):

(22) VP-internal PPs lack stress in presence of direct object

a. . . . dassthat

eina

Jxúnge

boy[einea

Gxéige

violinimin.the.DAT

Supermarktsupermarket

kaufte].bought‘. . . that a boy bought a violin in the supermarket.’

b. . . . dassthat

eina

Jxúnge

boy[einea

Gxéige

violinanto

einena.ACC

Freundfriend

schickte].sent‘. . . that a boy sent a violin to a friend.’

c. . . . dassthat

Marxía

Maria[K

xínder

childreninin

diethe.ACC

Schuleschool

fuhr].drove

‘. . . that Maria drove children to school.’

(23) VP-internal PPs are stressed in absence of direct object

a. . . . dassthat

eina

Jxúnge

boy[imin.a.DAT

Sxúpermarkt

supermarketlebte].lived

‘. . . that a boy lived in a supermarket.’

b. . . . dassthat

eina

Jxúnge

boy[nachto

Berlxín

Berlinfuhr].went

‘. . . that a boy went to Berlin.’

c. . . . dassthat

Marxía

Maria[inin

diethe.ACC

Schxúle

schoolfuhr].drove

‘. . . that Maria drove to school.’

The derivation of the major phrasing/stressing of the PPs in these two differentcases is given in (24):


(24) Spellout of VP-internal PP as a MaP/phrase stress depends on its po-sition in VP:

a. [ . . . [ object [PP verb


] ]VP ν ]νP →spellout

( objxect )MaP PP verb

b. [ . . . [ PP verb


]VP ν ]νP →spellout

(xPP)MaP verb

The happy prediction of (20) is that the VP-internal PP will get major phras-ing/stress only if it is the highest phrase in the spellout domain.

Finally, this theory makes correct predictions about the major phrasing/phrase stress of constituents in the spellout domains of higher phases. Con-sider the case of the subject of an all-new transitive sentence in German likethat in (1b). We assume that the subject might occupy a position within TP,either the specifier of ν or that of T. TP is the spellout domain of the CP phase,and the subject is thus the highest phrase in the spellout domain of the phaseand will be given major phrase status and phrase stress. Thus the sentence in(25) contains two spellout domains and on each of the domains the highestphrase will get major phrase status and phrase stress.

(25) Subject gets phrase stress in an all-new transitive sentence, as doesdirect object:CP[dass TP[ Maria vP[t1 VP[die Gesetze studiert

︸︷︷︸

]VP ν]vP

︸︷︷︸CP, νP phases: 2 spellout domains

]TP]CP →

Spellout: (Marxía)MaP

Maria(diethe

Gesxétze)MaP

lawsstudiert.is studying

‘Maria is studying the laws.’

The highest phrase account of prosodic spellout produces a sequence of ma-jor phrases, one for the highest phrase on each successive spellout domain. Inthis way, it predicts, correctly, the presence of as many major phrases (and ma-jor phrase stresses) as there are spellout-domain-internal highest phrases in aderivation.


Interestingly, if we followed the “simplest theory” of prosodic spellout andwere to assign a major phrase to the entire spellout domain, as proposed byIshihara (2007), the range of facts described above could not be accounted for.It is true, as Ishihara points out (p.c.), that if a transitive verb were universallyrequired to move out of VP on the ν-phase, Kahnemuyipour’s generalizationabout the necessity of phrase stress on the direct object could be derived byassuming a version of the simplest theory of prosodic spellout. But obligatoryverb movement would be of no help in deriving the fact that low direction-als and locatives within the VP do not receive phrase stress. The avoidance ofphrase stress in these cases requires a Highest Phrase Condition on prosodicspellout. This condition has, moreover, the advantage that it predicts the ap-pearance of phrase stress on any and all arguments that appear as highest phrasein the spellout domain of some phase. For the simplest theory to predict this, itwould have to assume that all such arguments themselves constitute phases andhence themselves constitute the domain for phrase stress. Indeed, Adger (2006)proposes that DP is a phase, and with this derives Kahnemuyipour’s general-ization. However, this proposal would still fail to account for the stresslessnessof low PPs following direct objects within VP in all-new sentences like thoseabove.5

Let us turn next to a consideration of the phrase stress patterns that wouldbe observed if it were not possible to satisfy the Highest Phrase Conditionwithin VP because the direct object is ineligible to receive phrase stress. Thiswould arise when the direct object is pronominal, for example. (26) containssentences from German illustrating these stress patterns. The direct object isthe weak indefinite was ‘something’.

(26) a. IchI

hab’have

gehört,heard

dassthat

MaríaMaria

[wassomething

gekáuftbought

hat].has

‘I’ve heard that Máría has bóught something.’b. Ich

Ihab’have

gehörtheard

dassthat

MaríaMaria

[wassomething

imin the.DAT

Súpermarktsupermarket

gekauftbought

hat].has

‘I’ve heard that María bought something in the súpermarket.’

In (26a) the presence of phrase stress on the embedded verb is related to theabsence of phrase stress on its pronominal direct object. (26b) shows that alow PP following a (stressless) pronominal direct object will now appear with

5. Truckenbrodt (2005) proposes that there is an interface principle Stress XP that is responsiblefor deriving patterns of phrase stress. This proposal runs into the same problem with VP-internal PPs in German that Adger’s does.


phrase stress, with the verb remaining stressless. The ineligibility of pronounsto receive phrase stress must be specified in the grammar. Note that the inabilityof phrase stress to appear on the pronoun, which is the highest phrase in thespellout domain, does not result in failure of prosodic spellout within VP. In(26b) the phrase stress appears on the next highest phrase in the VP, the lowPP. To account for this case, we might want to say that the prosodic spelloutprinciple (20) picks out the highest eligible phrase in the spellout domain forspellout as a major phrase:

(27) Spellout of low PP as a MaP/phrase stress following when direct ob-ject is pronoun:

[ . . . [ objectpro [PP verb


] ]VP ν ] ]νP →spellout

object (xPP)MaP verb

But the prosodic spellout of the verb in (26a) would not be derivable by theHighest Phrase Condition. Verb stress is in some sense the elsewhere case forprosodic spellout: if within the spellout domain there is no phrase available tospell out as major phrase, then the head gets prosodically spelled out.

(28) Elsewhere spellout of the verb as major phrase when other element(s)within VP are ineligible:

[ . . . [ objectpro verb


]VP ν ] ]νP →spellout

object (vxerb)MaP

The Elsewhere Condition for prosodic spellout could be formulated as in (29).

(29) The Elsewhere Condition on prosodic spelloutA spellout domain with eligible material must contain a major phrase.

Material eligible for spellout with major phrase is neither silent, a pronoun, orG-marked. A spellout domain consisting only of ineligible material will notcontain a major phrase at spellout. But as long as there is eligible material, amajor phrase will be spelled out. So a verb that is the sole eligible elementwithin a spellout domain will get major phrase stress.

In sum, this section has put forward a hypothesis concerning the conditionsunder which a constituent within a spellout domain may be spelled out as aprosodic major phrase and bearer of major phrase stress. The hypothesis con-sists of two conditions on prosodic spellout – (20), the Highest Phrase Condi-


tion (with its roots in Kahnemuyipour 2004), and (29), the Elsewhere Condi-tion. Together they specify that on any phase a spellout domain with eligiblematerial will contain exactly one major phrase. In Section 3.2 we will see thatthis theory of prosodic spellout, motivated here in connection with transitiveverbs, provides an account of patterns of phrase stress with intransitive verbsas well.

3.2. Intransitive verbs, topics, and the definition of phase heads

One of the classic puzzles we mentioned at the beginning is the contrast in verbstress/pitch accenting in the case of intransitive verbs in all-new sentences.

(30) a. IchI

hab’have

geträumt,dreamed

dassthat

derthe

RhéinRhine

ausgetrocknetdried.up

ist.is

‘I dreamt that the Rhíne dried up.’b. Ich

Ihab’have

gehört,heard

dassthat

Metállarbeitermetal workers

gestreiktgone.on.strike

haben.have‘I heard that métal workers went on strike.’

(31) IchI

hab’have

irgendwosomewhere

gelesen,read

dassthat

derthe

Königking

vonof

BáyernBavaria

spínnt.is.crazy‘I read somewhere that the King of Bavária was crázy.’

The subjects of the embedded intransitives in (30) and (31) bear phrase stress.The crucial observation is that in (31) the predicate also has to bear phrasestress, but in the examples of (30) no phrase stress on the predicates is neces-sary in all-new utterances: even without phrase stress, those predicates do nothave to be understood as given in the discourse. The embedded verb in (30a)is unaccusative, as shown by the fact that it takes the auxiliary sein (‘to be’)for its perfect forms. (30b) and (31) both have unergative verbs, as shown bythe fact that they form the perfect with the auxiliary haben (‘to have’). Thedifference between (30b) and (31) is usually seen as a contrast between even-tive (or stage-level) and stative (or individual-level) predicates. Supported bya battery of syntactic tests, Diesing (1990) argues that subjects of stage-levelpredicates and subjects of individual-level predicates do not occupy the samesyntactic positions. Subjects of stage-level predicates can appear in lower po-sitions than subjects of individual-level predicates. For example, in contrastto subjects of individual-level predicates, subjects of stage-level predicates are


acceptable to the right of certain particles and adverbials in German and canbe extracted from. According to Diesing, those positional differences are ul-timately also responsible for the observation of Carlson (1977) that bare plu-ral subjects of stage-level predicates can have existential interpretations, whilebare plural subjects of individual-level predicates can only have generic inter-pretations.

A central research question that came out of Diesing’s work concerns theconnection between subject positions and subject interpretations. Jäger (2001),who carefully reviews the large literature on the topic, reaches the conclusionthat a lexical property like the distinction between individual-level and stage-level predicates doesn’t seem to be directly responsible for the differentiatedsubject behavior Diesing identified. That behavior seems to be a rather indirectreflex of the syntactic representation of information structure. Jäger argues thatclauses must have syntactically represented topics, and subjects of stative (orindividual-level) predicates like be crazy, know, own and so on are often theonly phrases in their clause that could be topics. Since they are typically topics,bare plural subjects of individual-level predicates cannot easily be interpretedas existentials. From the current perspective, we might suspect that non-topicalsubjects occupy positions somewhere within TP, while topical subjects havemoved into a higher topic projection headed by a functional head Topic. ThisTopic projection could be part of an articulated periphery of the kind proposedin Rizzi (1997, 2004).

A major piece of support for Jäger’s proposal is that, under certain condi-tions, even “hard core” individual-level predicates can have low, non-topical,subjects. Those predicates can then be deaccented without being given in thediscourse and their bare plural subjects can have existential interpretations, oneof the main characteristics of low subjects according to Diesing. This phe-nomenon is illustrated by the examples in (32), which are modeled after En-glish examples from Fernald (2000). In both cases, the part in bold can beunderstood as all-new and the bare plural subjects can have existential inter-pretations.

(32) a. IchI

glaube,think

dassthat

inin

diesemthis.DAT

Baumtree

Áffenmonkeys

leben.live

‘I think that mónkeys live in this tree.’b. Ich

Iweiss,know

dassthat

diesesthis

HausGhouse

MaffiósiMaffiosi

besitzen.own

‘I know that Máffiosi own this house.’

If subjects of individual-level predicates don’t have to be topics, why can’tthe embedded subject of (31) be low and non-topical? That is, why can’t theembedded subject in (31) remain within TP and not move to the Spec position


of TopicP? And why is it that the bare plural subject of spinnen in (33) cannothave an existential interpretation, which according to Diesing shows that itmust be in a higher position?

(33) IchI

vermute,suspect

dassthat

Quácksalberquacks

spínnen.are.crazy

‘I suspect that quacks are crazy.’

According to Jäger, such facts follow from his requirement that every clausemust have a syntactically represented topic, that is, a position that is there tohost phrases denoting the individual or location that the main predication ofthe clause is about. In (32a), the topic is a location denoted by a PP, in (32b)the topic is an individual represented by a scrambled discourse-given object.Since the topic position is filled with other material, the topic requirement canbe satisfied even if the subjects remain within TP. With eventive predicates likedry up or go on strike the topic could also be a salient spatio-temporal locationrepresented by an unpronounced locative or temporal pronoun. (30a) and (30b)would then have implicit topics that might have been overtly expressed as in(34a) or (34b), for example:

(34) a. IchI

hab’have

geträumt,dreamed

dassthat

dannthen

derthe

RhéinRhine

ausgetrocknetdried.up

ist.is

‘I dreamt that then the Rhíne dried up.’b. Ich

Ihab’have

gehört,heard

dassthat

inin

BóchumBochum

diethe

Metállarbeitermetal workers

gestreiktgone.on.strike

haben.have

‘I heard that in Bochum the métal workers went on strike.’

The reason why the embedded subjects of (31) or (33) must be topical is thenthat, unlike claims about rivers drying up and metalworkers going on strike,claims about the permanent craziness of kings or quacks are not easy to un-derstand as claims about particular spatio-temporal locations: compare (35a),which may talk about some contextually salient spatio-temporal location, to(35b), which can only be about the Bavarian king himself.

(35) a. The king of Bavaria was getting dressed.b. The king of Bavaria was insane.

The only possible topics for (31) or (33) are then the king or the quacks, re-spectively, and consequently, the only possible candidates for topics in (31)or (33) are the subjects. On the account suggested by Jäger, there is still a


tight connection between subject positions and the individual-level/stage-leveldistinction, just as Diesing had claimed. But the connection isn’t as direct ase.g., Kratzer (1995) thought it was. The source of the syntactic differencesis the requirement that there must be a syntactically represented topic. Silentlocatives or temporal adverbials can satisfy the topic requirement with even-tive (stage-level) predicates, but such topics are often incompatible with stative(individual-level) predicates.

In what follows, we will show that the Topic status of what precedes an in-transitive verb has consequences for the phrase stress properties of the verb.6

Before we proceed, we have to clarify the phase status of Topic projections,however. In Chomsky’s work, phase heads are taken to be C and ν. There issome discussion in the literature whether ν is always a phase head, as arguedfor in Legate (2003), or whether ν should only count as a phase head if it in-troduces an external argument, which is the view expressed by Chomsky. Weare following Legate’s position here, which not only makes it possible to givea promising characterization of the inventory of possible phase heads, but isalso consistent with the prosodic facts, as we will see shortly. If there are Topicheads whose specifier positions represent the individual or location that themain predication of a clause is about, we have to ask ourselves whether theymight also be phase heads in addition to ν and C. Ultimately, whatever an-swer we give to this question will have to be justified by its theoretical success,but it would be good to have at least some initial rationale for why a partic-ular functional head should or should not be a phase head. One property thatmight characterize phase heads as a class might be the capacity to not merelyattract (“internally merge”), but introduce (“externally merge”) new material totheir specifier positions. The ν of transitive and unergative verbs introduces theverb’s external argument to its specifier position. The ν of unaccusative verbs,Topic heads, and C seem to have the capacity to introduce higher locatives,in particular framesetting locatives of the kind illustrated in (36) (Maienborn2001).

(36) . . . weilsince

inin

AustralienAustralia

Schwäneswans

schwarzblack

sind.are

‘. . . since in Australia swans are black.’

In contrast, heads related to tense or aspect, heads that attract focused con-stituents, and determiners never seem to genuinely introduce new material.Their specifier positions always seem to serve as potential landing sites formaterial that originates elsewhere. Assuming that the phase heads are precisely

6. Tokizaki (1999, 2006) also argues that the topic (categorial) status of subjects is fundamentalin determining the pitch accenting of the predicate.


those functional heads that, as a class, have the capacity for introducing newmaterial has implications that go beyond the data reviewed here. If higher ad-verbs are introduced by specialized functional heads, for example, as Cinque(1999) has proposed, and if the adverbs are in fact arguments of those heads, asMorzycki (2005) has argued, then functional heads introducing adverbs wouldhave to be phase heads on our account, with all the expected prosodic effects.7

Let us tentatively assume that Topic heads are phase inducing heads, then, andsee where this assumption takes us. What we expect in particular is that theTopic phase should involve an additional instance of spellout, on the TP com-plement of Topic, and that this will have consequences for the phrase stresspossibilities of intransitive verbs.

We are now in a position to apply our theory of spellout to examples like(30) to (32). The embedded clause of (30a) has an unaccusative verb. As hasoften been observed, the subjects of unaccusatives behave like direct objectsin many ways, and are commonly taken to be internal arguments of their verb(Perlmutter 1978). They should therefore originate within VP, and seem to beable to stay there in languages like German under certain conditions (Grewen-dorf 1989). If the case features of subjects and the tense features of verbs canin principle be valued via Agree, verbs and subjects of unaccusatives might notbe forced to leave their VP and may stay in situ. For (30a), we might then havean input configuration for spellout as shown in (37):

(37) Unaccusative subjects with no phrase stress on the verb

[TopicP pro TP[ [νP VP[der Rhein ausgetrocknet ist︸︷︷︸

Spellout domain of ν-phase

]VPν]νP T ]TP Topic] TopicP

Spellout: (derthe

Rhxéin)MaP

Rhineausgetrocknetdried up

istis

In (37), the topic position is filled with a silent locative pronoun representedas pro. There are three potential spellout domains in (37): TopicP (the com-plement of C), TP (the complement of Topic), and VP (the complement of

7. As Gussenhoven (1983) observed for English, the presence of an adverb intervening betweenthe subject and an intransitive verb forces the verb to bear pitch accent. Compare Máry ap-peared to Máry mystériously appéared/*appeared in an all-new utterance. This follows if theadverb is introduced by a phasal functional head.


ν). Within VP, the subject is the highest phrase, hence will receive the phrasestress. The verb complex remains unaccented. All material that is merged dur-ing later phases remains silent, hence is not eligible for phrase stress.

With unaccusative verbs, the spellout configuration remains unchanged ifthe VP also contains a directional phrase, for example, as in (38):

(38) IchI

hab’have

gehört,heard

dassthat

eina

Kíndchild

ausfrom

demthe.DAT

Zugtrain

gefallenfallen

ist.is

‘I heard that a chíld fell out of the train.’

The PP in (38) can be unaccented/unstressed without being given in the dis-course, and this is expected on our account. The only non-trivial spellout do-main remains VP in this case, and the subject, which is the highest phrasewithin VP, still receives the phrase stress. The PP and the verb cluster remainunaccented.

There is nothing to prevent unaccusative subjects from being topics, and thusassume a slightly different discourse role. Topics are not necessarily given inthe discourse, and this means that (39) and (40) are acceptable all-new utter-ances, too:

(39) IchI

hab’have

geträumt,dreamed

dassthat

derthe

RhéinRhine

aúsgetrocknetdried.up

ist.is

‘I dreamt that the Rhíne dried úp.’

(40) IchI

hab’have

gehört,heard

dassthat

eina

Kíndchild

ausfrom

demthe.DAT

Zúgtrain

gefallenfallen

ist.is

‘I heard that a chíld fell out of the tráin.’

In (39) and (40), the subject has moved into the Topic projection to achieve aparticular discourse effect, and thus receives phrase stress during the C-phase.The phrase stress/major phrase of the ν-phase now has to go to the highesteligible phrase in the VP. In (40), this is the directional PP. In (39), the verbcluster is the only non-silent material in the VP, and in the absence of a highestphrase, the Elsewhere Condition parses it as a major phrase. (Major phrasestress will fall on the participle because the auxiliary, a functional element, isnot parsed as a prosodic word.) In both cases, the spellout domain of the Topicphase, which is TP, does not contain any non-silent material that is not alreadycontained within VP, hence prosodic spellout on that TP can be ‘skipped’.


(41) Unaccusative subject with phrase stress on verb

[ C [TopicP der Rhein TP[t1 [νP t1 VP[t1 ausgetrocknet︸︷︷︸

ist]VP]νPν] T ]TP Topic

︸︷︷︸2 spellout domains

] TopicP ]C

Spellout: (derthe

Rhxéin)MaP

Rhine(

xaúsgetrocknet

dried upist)MaPis

(42) Unaccusative subject with phrase stress on PP

[ C [TopicP ein Kind TP[t1 [νP t1 VP[t1 aus dem Zug gefallen︸︷︷︸

ist]VP]νPν] T ]TP Topic


] TopicP ]C

Spellout: (eina

Kxínd)MaP

child(ausfrom

demthe

Zxúg)MaP

traingefallenfallen

istis

In our derivations so far, we have been assuming that finite verbs and verbclusters consisting of a participle and an auxiliary originate within the VP andcan be spelled out there. We may wonder, however, what guarantees that verbsand verb clusters are indeed spelled out within their VP in the cases we dis-cussed in this and the previous section. What would exclude a scenario where,in some way or other, the verb cluster ends up in a higher position – in ν or T,for example?

Finite verbs and auxiliaries move to T or C overtly in many languages. Assum-ing that movement is always local, that movement must have passed throughν. There is thus every reason to believe that finite verbs and auxiliaries canin principle move to ν, and movement to ν is therefore an option that shouldbe provided by Universal Grammar. Non-finite verb forms have been seen tomove, too. In languages like German or Dutch, verbs like to form clusters. Verbclusters consisting of two or more verbs have received a fair amount of attentionin the literature on West Germanic (see, in particular the discussion in Wurm-brand 2005). The order of verbs in a cluster can be quite variable, even within asingle language or dialect, suggesting movement. For example, in Dutch both(40a) and (b) are acceptable (Wurmbrand 2005: 228, Example (4)).


(43) a. . . . datthat

JanJan

hetthe

boekbook

gelezenread

heeft.has

‘. . . that Jan has read the book.’b. . . . dat Jan het boek heeft gelezen.

According to Wurmbrand, the predominant view on alternations like that il-lustrated in (43a) and (43b) is that one of the orders in (43) is basic, and theother one derived. There is, however, no agreement about the exact nature ofthe derivation. Whether the participle gelezen has overtly moved out of its VPin (43b), for example, possibly following earlier movement of the finite verb,is controversial. Boškovic (1997) argues that participles in Serbo-Croatian canmove out of VP and undergo head movement to a position below T. As il-lustrated in (44b), participles can optionally appear to the left of VP-externaladverbs in Serbo-Croatian (Boškovic 1997: 144, example (2)):

(44) a. JovanJovan

jeis

potpunocompletely

zaboravioforgotten

Petra.Peter

‘Jovan completely forgot Peter.’b. Jovan je zaboravio potpuno Petra.

It seems, then, that movement of participles out of VP to ν or T is also an op-tion that should be provided by Universal Grammar. In German, movement ofverb clusters to ν or T would be string vacuous in embedded sentences, how-ever, so the data we – or a language learning child for that matter – can relyon do not exclude the possibility that such movement does in fact take place.However, like any other kind of movement, verb movement should take placefor a reason. Everything else being equal, verbs should stay put and value theirfeatures via Agree. For verbs to move, there has to be a force to drive them.We want to propose that prosodic spellout economy may be a reason for short,string vacuous, verb movement in German. Sometimes, short verb movementcan have the effect that a whole spellout domain ends up silent, hence can infact be “skipped”. In such a case, we suggest, economy considerations forcethe verb to move, even though the movement is string vacuous. For the verb toremain all by itself in a spellout domain seems more costly than short move-ment to a higher spellout domain. The derivation of (30b) from above providesan illustration of such economy-driven verb movement.

In (30b), the subject does not have to be topical and can remain within TP,just as in (30a). But the verb is unergative in this case, rather than unaccusative,and this means that its subject originates in the specifier position of ν, henceoutside the spellout domain of ν. We now have to explain why it is possiblefor the verb cluster in (30b) to lack phrase stress without being given in thediscourse. The answer seems to be that this might be precisely the kind of


configuration where the verb cluster has to move into the ν-projection and thusout of the spellout domain of the ν-phase.

(45) Unergative subject with no phrase stress on the verb(stage-level predicate)

[TopicP pro TP[Metallarbeiter [νP t1 VP[gestreikt]VP gestreikt haben]νP] T︸︷︷︸

1 spellout domain

]TP Topic] TopicP

Spellout: (Metxállarbeiter)MaP

metal workersgestreiktgone on strike

habenhave

If verb movement leaves copies that are silent, moving the verb cluster in (45)makes it possible to literally skip a whole spellout domain and thereby servethe cause of economy. But to skip a spellout domain, the computational systemthat drives the syntax has to be able to register that a potential spellout domaindoes not in fact have to be sent to the phonological interpretation component.This skipping is possible, so it seems, if the spellout domain consists entirely ofpronouns, functional heads and leftovers of movement that the computationalcomponent can recognize – within the ν-phase where verb movement takesplace – as elements that will not have to be spelled out. Interestingly, silentelements, including PRO, traces, elided phrases, and unpronounced heads havealways played a central role in syntactic theory, and this suggests that the prop-erty of not needing to be spelled out is a property that is represented in thesyntax8. Whether or not material can remain silent depends in part on whetherits semantic content is recoverable. This is not a question that the phonologicalspellout component can be expected to answer. The decision about what can orcannot remain silent should thus be made in the syntax since that is the placethat mediates between phonological spellout and semantic interpretation. It istherefore plausible to assume that material that is to remain silent is markedas such in the syntax. It will then be possible for the computational compo-nent to recognize that a potential spellout domain that consists of nothing butelements that are to remain silent does not have to be sent to the phonologicalinterpretation component. The computational component might be guided byan economy principle like (46):

8. Merchant (2001) posits an E-feature to syntactically license ellipsis.


(46) Skipping prosodic spellouta. Skip prosodic spellout if possible.b. Prosodic spellout on a domain that consists of nothing but ele-

ments that are designated to remain silent can be skipped.

Returning to (45), within the ν-phase of the unergative verb, a decision canbe made in the syntax whether or not to move the verb cluster up to ν, basedon whether that movement will allow prosodic spellout to be skipped on thatphase. In (45) the verb cluster moves (together or one verb after the other) tosave on spellout. On the other hand, the verb cluster in (42) will not moveup to ν, precisely because this would not allow any spellout economy on theν-phase, since the PP has to be spelled out anyway. (41) now needs a slightlydifferent analysis from the one we gave earlier. In (41), moving the verb clusterto ν will make it possible to skip VP as a spellout domain. Assuming, as seemsmandatory in a phase-based system, that no look-ahead is allowed, movementof the verb cluster has to take place in this case, even though it now ends upas the only element in the spellout domain of Topic. Consequently, there isno “global” gain from moving the verb cluster, as long as it cannot move allthe way to Topic. This move, we suggest, is impossible on general grounds.Verbs do not agree with Topic heads in any way, and should therefore neverbe able to be attracted by Topic. The verb cluster in (38) has to be spelled outduring the Topic phase, then, and receives major phrasing (and stress) throughthe Elsewhere Condition, exactly as in our earlier analysis. A major differ-ence between unaccusatives and unergatives is expected if the VP contains PPslike directional phrases. If unergative subjects originate outside of VP in thespecifier position of ν, a VP-internal PP will receive phrase stress during theν-phase. We expect, then, that unergative verbs, directional phrases and otherPPs which lack phrase stress should always have to be interpreted as given inthe discourse. The expectation is borne out. (47b) contrasts with (38), in thatthe directional phrase in (47b) must be interpreted as given in the discourse.

(47) a. IchI

hab’have

gesehen,seen

wiehow

eina

Léhrerteacher

aufon

diethe

Úhrwatch

geschautlooked

hat.has

‘I saw a téacher look at his wátch.’b. #Ich

Ihab’have

gesehen,seen

wiehow

eina

Léhrerteacher

aufon

diethe

Uhrwatch

geschautlooked

hat.has

‘I saw a téacher look at his watch.’


(48) Unergative verb with directional PP

[TopicP pro TP[ [νP ein Lehrer VP[auf die Uhr geschaut hat︸︷︷︸

]VP] ννP] T


]TP Topic] TopicP

Spellout: (einA

Lxéhrer)MaP

teacher(aufon

diethe

xÚhr)MaPwatch

geschautlooked

hathas

In (48), the presence of the PP within VP makes it impossible to skip prosodicspellout on the spellout domain of the ν-phase by moving the verb cluster outof the VP. We end up with two instances of spellout, then, and with two majorphrases. Cases of lone unergative verbs like (45) do not show the expectedphrase stress in VP because of spellout economy and the forced movement ofthe verb (cluster) out of VP.

Turning to stative (or individual-level) unergative predicates as in (31), ifsentences must have topics, and a silent locative is not an option in such cases,the subject of (31) must be in the specifier position of Topic. The verb shouldmove up to ν to skip VP as a spellout domain. Being unable to move up allthe way to Topic, it has to be spelled out as the only element during the Topicphase and thus receives the phrase stress via the Elsewhere Condition.

(49) Unergative subject with phrase stress on verb (individual-level predi-cate)

[CPC[TopicPder König von Bayern TP[t1 [νP t1 VP[spinnt]VP spinnt νP

︸︷︷︸] T]TP Topic]


TopicP]CP

Spellout: (derthe

Königking

vonof

Bxáyern)MaP

Bavaria(sp

xínnt)MaP

is crazy

Let us now look at the two sentences in (32), which both have stative (individual-level) predicates. By filling the obligatory topic position with other material,they both allow the subject to remain within TP, possibly even in the positionwhere it originated. In (32a), a PP occupies the specifier of Topic, and in (32b)the direct object (which has to be given in this case) has been moved there. Thephrase stress assigned during the Topic phase will then fall on the subject asthe highest phrase within TP. The verb can remain unaccented without havingbeen discourse given, provided it moves out of the VP and adjoins to ν. Our


hypothesis of economy-driven verb movement predicts this movement shouldtake place, since it would evacuate the VP, and make it possible to skip aninstance of prosodic spellout.

(50) Unergative subject without phrase stress on verb (individual-level pred-icate)[C [TopicP in diesem Baum TP[[νP Affen VP[leben]VP leben]νP] T

︸︷︷︸]TP Topic]


TopicP ]C

Spellout: (inin

diesemthis

Bx

áum)MaPtree

(xÁffen)MaPmonkeys

lebenlive

(51) Transitive sentence with scrambled object9 (individual-level predi-cate)

[C [TopicP dieses HausG TP[[νP Maffiosi VP[besitzen]VP besitzen]νP] T︸︷︷︸

]TP Topic] TopicP


]C

Spellout: diesesthis

Haushouse

(Maffixósi)MaP

Maffiosibesitzenown

If the verb were left within the VP in (32a) and (32b), it would have to receivephrase stress during the ν phase, and the result would be (52a) and (52b):

(52) a. #IchI

glaube,think

dassthat

inin

diesemthis

Baumtree

Áffenmonkeys

lében.live

‘I think that mónkeys líve in this tree.’b. #Ich

Iweiss,know

dassthat

diesesthis

HausGhouse

MaffiósiMaffiosi

besítzen.own

‘I know that Máffiosi ówn this house.’

It is hard to understand the embedded verbs of (52a) and (52b) as non-contrast-ive. Our account predicts this, in that the only way the verbs could receivephrase stress here is if the verb were F-marked and subject to the ContrastiveFocus Prominence Rule.

The final case to discuss involves the stressing of the verb with an adjacentG-marked given direct object that was observed in Section 2.1. As we have just

9. This example shows that the givenness of a direct object does not entail the stressing of theverb, as one might be led to think on the basis of the sentences in (6)–(7) in Section 2.1.


seen in the Maffiosi sentence above, a given direct object can move to a higherTopic position. In (53b), too, the direct object is likely to have moved out of itsVP, this time to a position below the subject. That given definite objects liketo leave their VP is shown by the grammaticality contrast between (53b) and(53c), where we are using negation to mark the left edge of VP.

(53) a. IchI

hab’have

gehört,heard

dassthat

Marxía

Mariadiethe

GesetzeGlaws

studxíert.

is studying‘I’ve heard that Mária is stúdying the laws.’

b. Ich hab’ gehört, dass Marxía die GesetzeG nicht stud

xíert.

c. ?Ich hab’ gehört, dass Marxía nicht die GesetzeG stud

xí ert.

In (53ab), where the subject is not itself necessarily a Topic, the direct objectmight be in a lower Topic position, assuming that there can be multiple Topicpositions (Rizzi 1997, 2004). Indeed, given our theory, we would need to as-sume the presence of a Topic-phase in the case of the moved direct object in(53), in order to derive the phrase stress on the verb. Movement of the objectout of the VP forces verb movement to ν on our account, and the spellout do-main for the ν-phase can then be skipped. The verb would as a consequence bethe only element in the spellout domain of the lower Topic phrase and receivephrase stress at that point by the Elsewhere Condition.

3.3. Summary

Phase-based spellout has allowed us to make sense of the distribution of ma-jor phrase stress on verbs and phrases in German. A single major phrase isintroduced for any phasal spellout domain containing eligible material (mate-rial that is not G-marked, a function word, or silent). Verb movement out ofVP appears to be driven by an economy principle that favors skipping prosodicspellout on a phase, if that is possible. Modulo such spellout skipping, there isa direct relation between the number of phases and the number of major phrasestresses in a sentence. Key to our understanding of the necessity of phrasestress on stative (individual-level) verbs has been the idea that the topic statusof the subject, which is independently motivated, implies the presence of anadditional spellout domain, produced by a Topic-phase.10

10. Topics that are “dislocated” at the periphery and leave a copy pronoun in the sentence have aneven more dramatic effect on sentence prosody, in that they appear to be parsed in a separateintonational phrase from the body of the sentence, as shown for Greek by Revithiadou and


The Highest Phrase Condition for prosodic spellout says that within thespellout domain of a phase, only the highest eligible phrase will get majorphrasing (and stress). The consequence is that any non-highest phrases withinthe VP will not receive default phrase stress even when discourse-new, sinceVP is the lowest spellout domain of the sentence. By contrast, eligible phrasesthat appear in VP-external position, always in specifier positions of higher pro-jections, can all receive phrase stress, if they are the highest phrase in the spell-out domain of some phase. This is true of phrases that move up out of VP andof phrases that are generated in higher specifier positions. Amongst phrasesthemselves, bearing phrase stress is a function of position within a spelloutdomain. As for verbs, only in the absence of eligible highest phrases within aspellout domain will the presence of major phrase stress on an eligible verb bederived – through the Elsewhere Condition. These principles can result in all-new sentences in German where a succession of phrase stresses on specifiersis followed by a VP where, after the first stressed phrase, a potentially longstretch of non-phrase-stressed material is found.

As a final reminder, the G-marking of a discourse-given constituent rendersit ineligible for major phrasing and stress, and can therefore lead to differentpatterns of major phrasing and stress from those that the Highest Phrase andElsewhere Conditions would predict in all-new sentences. In addition, the F-marking of a contrastive focus constituent may produce a phrase stress thatwould not be predicted by these two principles of default major phrasing andstress. The prosodic organization of a sentence is a function of both its con-stituency in terms of phases and its semantic/pragmatic properties, as repre-sented by F-marking and G-marking.

4. The phonological interpretation of prosodic spellout structure

In the introduction to this article we suggested that the distribution of defaultpitch accents in German and English is a function of the default phrase stresspatterns defined in these languages. In subsequent sections we suggested thatthese default phrase stress patterns were themselves a function of a prosodicphrasing that is assigned as part of the phase-based spellout of syntactic struc-ture. The formulation of prosodic spellout that we offered above states that thehighest phrase in the spellout domain of a phase is spelled out as a correspond-ing prosodic constituent in phonological representation, which is equated withthe prosodic major phrase (cf. (20), and that in any case a spellout domaincontaining eligible material will contain a major phrase (cf. (29)). Our hypoth-esis, therefore, is that default pitch accent distribution in English and German

Spyropoulos (2004), who propose a minimalist account of this phrasing.


is ultimately a reflex of principles that relate syntactic constituency to prosodicconstituency. The goal of this section is to show that the theory of phase-basedsyntactic-constituency-sensitive spellout can indeed provide the basis for aninsightful account of pitch accent distribution in these languages.

4.1. Phonological constraints on prosodic parsing

We have seen that prosodic spellout as defined above produces a prosodic struc-ture in which the verb may fail to be parsed into a major phrase. Function wordslike the determiner die and the complementizer dass in German also show anincomplete prosodic parsing. This is illustrated by the transitive sentences be-low in (54).

(54) Output of universal principles of prosodic spellout (provisional)11

a.

intonational phrase

major phrase

prosodic word

(((Mari

))a

Maria

(studier

)t

is studying

(

diethe

(Gesetz

)))e.

laws

b.

intonational phrase

major phrase

prosodic word

(

. . . dassthat

((

Mari

))a

Maria

(

diethe

(Gesetz

))e

laws

(studier

)

)t.

is studying

Intonational phrase and prosodic word, which we assume are produced by in-dependent principles of spellout,12 are also represented here. The verb, moreprecisely the prosodic word to which we assume it corresponds, is parsed aspart of the intonational phrase, but not as part of a major phrase. From thepoint of view of phonological theory, such structures would be consideredto be marked. An exhaustive parsing of the string dominated by a prosodicconstituent into prosodic constituents at the next level down in the prosodichierarchy is arguably the ideal for phonological representation (as argued in

11. Sections 4.2 and 4.3 argue that the full representation includes the level of minor phrase aswell. The minor phrase could be understood as the spellout of any syntactic branching beneaththe level of major phrase (cf. Kubozono 1993).

12. In the perspective of the current article, in which designated syntactic constituents are spelledout as corresponding prosodic constituents, prosodic word could be understood as the spelloutof lexical (not functional) heads, while intonational phrase could be the spellout of ‘commaphrase’ (cf. Potts 2005; Selkirk 2005), an instance of which would be the Force Phrase (Rizzi1997), the highest node of the sentence.


Selkirk 1995 and elsewhere). We might therefore expect outputs of prosodicspellout like (54ab) to be altered in order to conform more closely to prosodicstructure ideals. One sort of alteration would simply promote the verb to majorphrase status, but this is not observed, perhaps because of the increase in majorphrase heads that would result. A different type of alteration to spellout struc-ture would involve adjunction of a “stray” verb to an adjacent major phrase,creating a nested major phrasing, as seen in (55):

(55) Prosodic markedness-driven alterations to output of prosodic spelloutprinciples

a.

intonational phrase

major phrase

prosodic word

(((

Mari

))a

Maria

((studier

)t

is studying

((diethe

Gesetz

))))e.

laws

b.

intonational phrase

major phrase

prosodic word

(

. . . dassthat

((

Mari

))a

Maria

((

diethe

(Gesetz

))e

laws

(studier

)))t.

is studying

We will see below that adjoined major phrase structures like these, which con-tain a single major phrase head for nested major phrases, provide the basis forassigning the sort of phrase stress patterns that are needed to explain the dis-tribution of default pitch accents in English and German.13 And so we thinkit is worthwhile to entertain a theory of prosodic phrasing which includes uni-versal interface principles of prosodic spellout as sketched above, and, as partof the phonology, an optimality-theoretic ranking of prosodic markedness con-straints, which operate to produce surface prosodic structures that are morenearly phonologically ideal. With such a theory, language-particular variationin prosodic phrasing would be the consequence of the phonology: differentlanguage-particular rankings of prosodic markedness constraints could giverise to different alterations to the prosodic structure produced by the universalprosodic spellout principles. For example, it is conceivable that in some otherlanguage, a verb that is in configuration (54a) might adjoin to the major phrasethat precedes it, rather than to the major phrase that follows. Revithiadou andSpyropoulos (2004) report on work showing that this is true in Greek. As for

13. Clearly, it would be desirable to have independent evidence for this nested structure from otherphonological or phonetic phenomena, including the possible presence of boundary tones, orthe blocking of assimilatory phenomena that the presence of the internal major-phrase edgewould imply.


where the verb is adjoined in the general case, one might wonder whether thereis a tendency for languages with major phrase-final stress placement like Ger-man to favor the “proclitization” of the verb, while major-phrase-stress-initiallanguages might favor the “encliticization” of the verb. All this needs to beexplored.

This article is not the place to develop a sustained argument in favor of thenew claim that cross-linguistic variation in default prosodic phrasing is a func-tion of language-particular differences in the phonological component of thegrammar, not of differences in the syntax-phonology interface, namely spell-out. Rather, we must content ourselves in the last two sections with merelyshowing that major phrase structures like those in (55) are needed in an ac-count of default pitch accenting in German and the phrase stress on which it isbased, and with the observation made above that phonological pressures couldplausibly give rise to the structures in (55) on the basis of the phrasings in(54), which we have hypothesized are the output of universal prosodic spelloutprinciples.

4.2. Default pitch accent as tonal enhancement of prosodic headedness/stress

Standard phonological theory distinguishes the representation of tone from therepresentation of prosodic structure prominence (headedness) and is chargedwith providing an account of the interdependence between the two (see Yip2002). The descriptive term “pitch accent”, in the sense which we have beenusing it, refers to a tonal entity which is confined to a “culminating” positionwithin some prosodic domain, often characterized as a position of stress. Inso-called “intonational languages” like English, Dutch and German, unlike inlexical tone languages, the tonal elements of the pitch accents are not part ofthe underlying representations of morphemes with additional segmental con-tent. Rather, in an intonational language it is only in surface representationthat a pitch accent comes to be associated with a prosodically prominent syl-lable in a word of the sentence. There appear to be two distinct sources ofpitch accents in intonational languages – default pitch accents, which are non-meaning-bearing tones epenthesized in the phonology, and pitch accents thatare tonal morphemes. In both cases, these tonal elements end up located on asyllable with local prosodic prominence.

In English, Dutch and German, a variety of pitch accent types have been iso-lated (Grice, Baumann and Benzmüller 2005 for German; Gussenhoven 1983,2003 for Dutch; Beckman and Ayers-Elam 1997 for English), some of whichare claimed to make distinctive meaning contributions to the sentence (see, forexample, Pierrehumbert and Hirschberg 1990 on English). It has been proposedby Büring (1997), for example, that the contrastive topic in German is marked


by a L*+H pitch accent. Conceivably, that L*+H pitch accent is the realizationof contrastive topic morphology, produced as part of the spellout of the con-trastive topic phrase, and then located on the main stressed syllable within it.But there are other pitch accents – in English, Dutch, German, and other lan-guages – which have no particular meaning contribution. These pitch accentsappear to be epenthetic tonal elements whose function, if any, is to enhancethe head of a prosodic constituent. This class of default pitch accents includesthose that are introduced on the main stress of every word in Cairene Ara-bic (Hellmuth 2006)14, on the initial (head) syllable of every accentual/minorphrase in Korean (Jun 1995), and on the main stress of what has been referredto as phonological phrase in Bengali (Hayes and Lahiri 1991). Quite conceiv-ably, tonal elements may in principle be introduced by default on a head/mainstress prominence at any level of the prosodic hierarchy, and languages willdiffer depending on whether and at what level of prosodic headedness such atonal epenthesis is defined. So where do the default epenthetic pitch accents ofEnglish and German fit in?

The paradigm of pitch accent distribution which we gave for all-new transi-tive sentences at the beginning of this article was as follows (the schematic T*stands for the tone(s) comprising the default pitch accents, which may vary inquality from one dialect of German to another):

(56) a. MariT*

a studie(T*)

rt die GeseT*

tze.

‘Maria is studying the laws.’b. . . . dass Mari

T*a die Gese

T*tze studie

–rt.

‘. . . that Maria is studying the laws.’

In sentences like these, a pitch accent necessarily falls on the subject and ob-ject arguments; the verb necessarily lacks accent when it is sentence-final, butshows optional accent if it precedes an object. Which level of phrase stress is itthat calls for the presence of pitch accent in German? The necessary presenceof pitch accent on the DPs suggests it might be major phrase stress, but thevagaries in the pitch accenting of verbs rules against this. For a medial verb toreceive optional pitch accent, it would have to be optionally promoted to majorphrase status and thereby receive major phrase stress and an epenthetic pitchaccent. And we would then have to stipulatively block the promotion of finalverbs to major phrase stress status, since the option for pitch accenting is notavailable finally. Another argument against the promotion of nonfinal verbs tomajor phrase status comes from English, where it has been shown that pitch

14. Default pitch accent in Nubi (Gussenhoven 2006) is also conceivably a case like this.


accented verbs in all-new, neutral sentences fail to be followed by the L- pe-ripheral tone that should be found if the verb itself constituted a major phrase(Selkirk 2002; Katz and Selkirk 2006).

A more promising solution is to (i) assume the existence of the lower-levelminor phrase in English and German, as in Japanese and Korean, (ii) assumethat a level of minor phrase headedness is defined and (iii) to restrict the epen-thesis of default pitch accent in English and German to the head (main stress)of the minor phrase.

(57) The Pitch Accent Constraint (English, German)The head of a minor phrase requires a tonal pitch accent.

Because a major phrase stress will always also be a minor phrase stress, giventhe nature of the phonological representation of stress, (57) immediately pre-dicts the necessary presence of pitch accent on elements that bear major phrasestress. As for the categorical absence of pitch accents, or their optional pres-ence, Section 4.3 is devoted to examining the grammatical system that wouldgovern the distribution of those minor phrase stresses (and pitch accents) whichdo not coincide with major phrase stress in the sentence.

4.3. Distribution of minor phrase stress in English and German

Let’s consider first the prosodic structure representation of the verb-final sen-tence amplified with the level of minor phrasing as in (58). The verb would notbe parsed as a minor phrase, would lack minor phrase stress, and hence bearno pitch accent.

(58) Grammatical minor phrasing/pitch accenting with final transitive verb

major phrase

minor phrase

prosodic word

OK

((

. . . dass(

Mar

xxxí

T*

)))a

(((

die(Ges

xxxé

T*tz

)))e

(stud

xier

)

)t

‘. . . that Maria is studying the laws.’

Prosodic phonology has a ready explanation for why minor phrase stress/pitchaccent should be lacking on the final verb. In German and English a prosodicconstraint, call it Head-EdgeMaP-R, calls for major phrase stress to be right-most within the major phrase (cf. (13)).15 We can understand this constraint

15. See Ishihara and Féry (2006) for a similar account of post stress deaccenting in German.


as requiring that major phrase stress fall on the rightmost minor phrase withinmajor phrase (cf. McCarthy 2003 on edgemostness in locating prosodic heads).We also assume that this constraint holds of both major phrases in a nestedstructure such as (58), which are headed by the same major phrase stress. (58)satisfies the Head-EdgeMaP-R constraint on the larger major phrase containingthe verb because there is no minor phrase lying between the major phrase stresson Gesetze and the right edge of the major phrase. (59) illustrates the fact that ifthere were a minor phrase stress (and accompanying pitch accent) on the verb,Head-EdgeMaP-R would be violated.

(59) Ungrammatical minor phrasing/pitch accenting with final transitiveverb

Head-EdgeMaP-R rules this out

*

major phrase

minor phrase

prosodic word

((

. . . dass

that

(Mar

xxxí

T*

)))a

Maria

(((

die

the

(Ges

xxxé

T*tz

)))e

laws

((stud

xxieT*

r

)))t

is studying‘. . . that Maria is studying the laws.’

So it seems we have a simple standard sort of phonological explanation for thecategorical lack of pitch accent on a final transitive verb in an all-new sentence.Head-EdgeMaP-R must simply outrank any other grammatical constraint whichwould seek to introduce a minor phrase in that position.

This same account predicts the asymmetry observed in the distribution ofpitch accents: words that appear to the left of major phrase stress within themajor phrase will not be prevented by Head-EdgeMaP-R from having status asheads of minor phrase, and so should be able to show pitch accents. Theseinclude the cases of medial transitive verbs, for example, if we adopt the as-sumption, illustrated in (60a), that they are adjoined to the major phrase on theirright. If these medial verbs were, by contrast, adjoined to the major phrase onthe left, they would be incapable of bearing minor phrase stress, just like theverbs in sentence-final position. The representations in (60ab) show the twooptions in minor phrase organization that we propose for the cases of presenceand absence of pitch accent on medial verbs:


(60) Options in minor phrasing and pitch accenting of medial verbs

a.

major phrase

minor phrase

prosodic word

(((Mar

xxxi

T*

)))a

(((stud

xxieT*

r

))t

(((die Ges

xxxe

T*tz

))))e.

b.

major phrase

minor phrase

prosodic word

(((Mar

xxxi

T*

)))a

(

(stud

xier

)t

(((die Ges

xxxe

T*tz

))))e.

‘Maria is studying the laws.’

With this, we hope to have provided some initial plausibility for our hypothesisthat constraints governing the presence/absence of minor phrasing and minorphrase stress are the source of an explanation for the distribution of defaultpitch accents in German and English. The minor phrase has been most thor-oughly motivated and examined in Japanese and Korean. Further argumenta-tion for positing the minor phrase and its role in pitch accenting in German andEnglish is of course necessary; it must include an investigation of the prosodicspellout principle(s) that would introduce minor phrasing, as well as purelyphonological principles that might govern minor phrase organization as well.16

This remains a project for future research.

4.4. Where is sentence phonology?

In construing constraints on the syntax-phonology interface as part of phase-based spellout, as we have in this article, the question naturally arises as tohow much of phonology (and phonetics) is done during spellout. The prosodicadjunction of the stray verb, the determination of optional minor phrasing, andthe epenthesis of default pitch accents discussed above could in fact all wait tillthe syntactic derivation and its multiple spellout was complete. Conceivably,during spellout only a partial phonological representation would be defined,precisely that which allows satisfaction of the interface constraints on prosodicphrasing and stress – namely the Highest Phrase and Elsewhere Conditionson major phrasing; constraints on intonational phrasing, minor phrasing andprosodic words; and the G-marking- and F-marking-sensitive constraints De-stress Given and Contrastive Focus Prominence Rule. The incomplete prosodic

16. Selkirk (2006a) includes a more detailed discussion of optional pre-major phrase pitch ac-centing in English.


structure representation that is the cumulative result of potentially multiplespellout(s) on the whole sentence could be the input to the actual phonolog-ical component – say an optimality-theoretic constraint ranking – which woulddefine a full surface phonological representation, and provide the input to thephonetics.

A grammar with a post-syntactic phonological component would give a re-stricted role to the syntactic derivation in determining sentence phonology,seeing the effect of syntax on phonology (and phonetics) as mediated by itseffect on prosodic constituency and stress, as in earlier models of the syntax-phonology interface (e.g., Selkirk 1986; Nespor and Vogel 1986; Truckenbrodt1995). Further research needs to investigate whether the phonological compo-nent should be fully integrated into phase-based spellout, where it could pro-duce opaque “cyclic” effects not capturable by a post-syntactic phonologicalinterpretation.

University of Massachusetts Amherst

References

Adger, David (2006). Stress and phasal Syntax. Ms., Queen Mary College, University of London.Beckman, Mary and Gayle Ayers-Elam (1997). Guidelines for ToBI labelling. Ms., Ohio State

University. Version 1997. http://www.ling.ohio-state.edu/~tobi/Beckman, Mary and Janet Pierrehumbert (1986). Intonational structure in English and Japanese.

Phonology Yearbook 3: 255–310.Boškovic, Željko (1997). The Syntax of Nonfinite Complementation. Cambridge, MA: MIT Press.Bresnan, Joan (1971). Sentence stress and syntactic transformations. Language 47: 257–281.Büring, Daniel (1997) The Meaning of Topic and Focus – The 59th Street Bridge Accent. London:

Routledge.Carlson, Gregory (1977). Reference to kinds in English. Unpublished Ph.D dissertation, University

of Massachusetts at Amherst.Chomsky, Noam (2000). Minimalist inquiries: The framework. In Step by Step: Essays in Mini-

malist Syntax in Honor of Howard Lasnik, R. Martin, David Michaels, and Juan Uriagereka(eds.), 89–115. Cambridge, MA: MIT Press.

— (2001). Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.),1–52. Cambridge, MA: The MIT Press.

— (2005a). Three factors in language design. Linguistic Inquiry 36: 1–22.— (2005b). On phases. Ms., MIT Department of Linguistics.Cinque, Guglielmo (1993). A null theory of phrase and compound stress. Linguistic Inquiry 24:

239–297.Cinque, Gugliemo (1999). Adverbs and Functional Heads – A Crosslinguistic Perspective. Oxford:

Oxford University Press.Diesing, Molly (1990). Indefinites. Cambridge, MA: MIT Press.D’Imperio, Mariapaola, Gorka Elordieta, Sónia Frota, Pilar Prieto and Marina Vigário (2005).

Intonational phrasing in Romance: the role of syntactic and prosodic structure. In Prosodies(With Special reference to Iberian Languages), Sónia Frota, Marina Vigário and Maria JoãoFreitas (eds.), 59–98. Berlin: Mouton de Gruyter.

Fernald, Ted (2000). Predicates and Temporal Arguments. Oxford: Oxford University Press.


Féry, Caroline and Vieri Samek-Lodovici (2006). Focus projection and prosodic prominence innested foci. Language 82: 131–150.

Frota, Sonia (2000). Prosody and Focus in European Portuguese: Phonological Phrasing and In-tonation. New York: Garland.

German, James, Janet Pierrehumbert and Stefan Kaufmann (2006). Evidence for phonologicalconstraints on nuclear accent placement. Language 82: 151–168.

Grewendorf, Günther (1989). Ergativity in German. Dordrecht: Foris.Grice, Martine, Stefan Baumann, and Ralf Benzmüller (2005). German Intonation in

autosegmental-metrical phonology. In Prosodic Typology: The Phonology of Intonation andPhrasing, Sun-Ah Jun (ed.), 55–83. Oxford: Oxford University Press.

Gussenhoven, Carlos (1983). Focus, mode, and the nucleus. Journal of Linguistics 19: 377–417.— (1992). Sentence accents and argument structure. In Thematic Structure: Its Role in Grammar,

Iggy Roca (ed.), 91–106. Berlin, New York: Foris Publications.— (2003). ToDI (Transcription of Dutch Intonation), 2nd ed. http://todi.let.kun.nl/ToDI/home.

htm.— (2006). Between stress and tone in Nubi word prosody. Phonology 23 (2): 193–224.Halle, Morris and Jean-Roger Vergnaud (1987). An Essay on Stress. Cambridge, MA: MIT Press.Hayes, Bruce (1981). A metrical theory of stress rules. Unpublished Ph.D. dissertation, MIT, Cam-

bridge.— (1995). Metrical Stress Theory: Principles and Case Studies. Chicago: The University of

Chicago Press.Hayes, Bruce and Aditi Lahiri (1991). Bengali intonational phonology. Natural Language and

Linguistic Theory 9: 47–96.Hellmuth, Sam (2006). Intonational pitch accent distribution in Egyptian Arabic. Unpublished PhD

dissertation, SOAS, University of London.Inkelas, Sharon and Draga Zec (eds.) (1990). The Phonology–Syntax Connection. Chicago: The

University of Chicago Press.Ishihara, Shinichiro (2007). Major phrase, focus intonation and multiple spellout. This issue.Ishihara, Shin and Caroline Féry (2006). Phonetic Correlates of Second Occurrence Focus. In Pro-

ceedings of the 36th North Eastern Linguistic Society (NELS 36), Christopher Davis, Amy-Rose Deal and Youri Zabbal (eds.), 371–384. Amherst, MA, GLSA.

Jackendoff, Ray (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MITPress.

Jacobs, Joachim (1988). Fokus-Hintergrund Gliederung und Grammatik. In Intonationsforschun-gen, Hans Altmann (ed.), 89–134. Tübingen: Niemeyer.

— (1993). Integration. In Wortstellung und Informationsstruktur, Marga Reis (ed.), 64–116. Tub-ingen: Niemeyer.

Jäger, Gerhard (2001). Topic–Comment structure and the contrast between stage-level and individ-ual-level predicates. Journal of Semantics 18: 83–126.

Jun, Sun-Ah (1995). The Phonetics and Phonology of Korean Prosody. New York: Garland.— (1998). The Accentual Phrase in the Korean prosodic hierarchy. Phonology 15(2):189-226.— (2005). Prosodic Typology: The Phonology of Intonation and Phrasing. Oxford: Oxford Uni-

versity Press.Kahnemuyipour, Arsalan (2003). Syntactic categories and Persian stress. Natural Language &

Linguistic Theory 21: 333–379.— (2004). The syntax of sentential stress. Unpublished Ph.D dissertation, University of Toronto.Katz, Jonah and Elisabeth Selkirk (2006). Pitch and duration scaling for contrastive focus: a phrase

stress analysis. Ms., University of Massachusetts Amherst.Kiss, Katalin É. (1998). Identificational focus versus information focus. Language 74 (2): 245–

273.Kratzer, Angelika (1995). Stage-level and individual-level predicates. In The Generic Book, Gre-

gory N. Carlson and Jeffrey F. Pelletier (eds), 125-175. Chicago: Chicago University Press.


Kratzer, Angelika (2004). Interpreting focus: Presupposed or expressive meanings? TheoreticalLinguistics (Special issue on Interpreting Focus) 30 (1): 123–136.

Krifka, Manfred (1992). A compositional semantics for multiple focus constructions. In Informa-tionsstruktur und Grammatik, Special issue of Linguistische Berichte, Joachim Jacobs (ed.),17–53. Opladen: Westdeutscher Verlag.

Kubozono, Haruo (1993). The Organization of Japanese Phonology. Tokyo: Kurosio Publishers.Ladd, D. Robert (1980). The Structure of Intonational Meaning: Evidence from English. Bloom-

ington: Indiana University Press.— (1996). Intonational Phonology. Cambridge: Cambridge University Press.Legate, Julie Anne (2003). Some interface properties of the phase. Linguistic Inquiry 34: 506–516.Maienborn, Claudia (2001). On the position and interpretation of locative modifiers. Natural Lan-

guage Semantics 9: 191–240.McCarthy, John J. (2003). OT constraints are categorical. Phonology 20:75–138.McGinnis, Martha (2001). Variation in the phase structure of applicatives. Linguistic Variation

Yearbook 1: 105–146.Merchant, Jason (2001). The Syntax of Silence. Sluicing, Islands, and the Theory of Ellipsis. Ox-

ford: Oxford University Press.Morzycki, Marcin (2005). Mediated Modification: Functional Structure and the Interpretation of

Modifier Positions. Unpublished Ph.D. dissertation, University of Massachusetts Amherst.Nespor, Marina and Irene Vogel (1983). Prosodic structure above the word. In Prosody: Models

and Measurements, Anne Cutler and D. Robert Ladd (eds.), 123–140. Heidelberg: Springer.— (1986). Prosodic Phonology. Dordrecht: Foris Publications.Perlmutter, David (1978). Impersonal passives and the unaccusative hypothesis. In Proceedings

of the Fourth Annual Meeting of the Berkeley Linguistics Society, 157–189. University ofCalifornia Berkeley: Berkeley Linguistics Society.

Pierrehumbert, Janet and Mary Beckman (1988). Japanese Tone Structure. Cambridge, MA: MITPress.

Pierrehumbert, Janet B. and Julia Hirschberg (1990). The meaning of intonational contours in theinterpretation of discourse. In Intentions in Communication, Philip R. Cohen, Jerry Morgan,and Martha E. Pollack (eds.), 271–311. MA: MIT press.

Poser, William J. (1984). The phonetics and phonology of tone and intonation in Japanese. Unpub-lished Ph.D. dissertation, MIT, Cambridge.

Potts, Christopher (2005). The Logic of Conventional Implicatures. Oxford Studies in TheoreticalLinguistics. Oxford: Oxford University Press.

Prieto, Pilar (2005). Syntactic and eurhythmic constraints on phrasing decisions in Catalan. StudiaLinguistica 59 (2/3): 194–222.

Revithiadou Anthi, and Vassilios Spyropoulos (2004). The multiple spell-out hypothesis and thephonological component: evidence from Greek. In Proceedings of the 35th Meeting of theNortheast Linguistics Society (NELS 35), Leah Bateman and Cherlon Ussery (eds.), 523–538.University of Massachusetts Amherst: GLSA.

Rizzi, Luigi (1997). The fine structure of the left periphery. In Elements of Grammar: Handbookof Generative Syntax, Liliane Haegemann (ed.), 281-337. Dordrecht: Kluwer.

— (2004). Locality and left periphery. In Structures and Beyond. The Cartography of SyntacticStructures, Volume 3, Adriana Belletti (ed.), 223–251. Oxford: Oxford University Press.

Rooth, Mats (1992). A theory of focus interpretation. Natural Language Semantics 1: 75–116.— (1996). Focus. In The Handbook of Contemporary Semantic Theory, Shalom Lappin (ed.),

271–297. London: Blackwell.Schmerling, Susan (1976). Some Aspects of English Sentence Stress. Austin TX: University of

Texas Press.Seidl, Amanda (2001). Minimal Indirect Reference: A Theory of the Syntax-Phonology Interface.

London: Routledge.


Selkirk, Elisabeth (1978/1980). On prosodic structure and its relation to syntactic structure. InNordic Prosody 2, T. Fretheim (ed.), 128–149, Trondheim: TAPIR.

— (1980). The role of prosodic categories in English word stress. Linguistic Inquiry 11 (3):563–605.

— (1984). Phonology and Syntax: The Relation between Sound and Structure. Cambridge MA:MIT Press.

— (1986). Derived domains in sentence phonology. Phonology Yearbook 3: 371–405.— (1995). Sentence prosody: intonation, stress and phrasing. In The Handbook of Phonological

Theory, John Goldsmith (ed.), 550–569. London: Blackwell.— (2002). Contrastive FOCUS vs. presentational focus: Prosodic evidence from right node rais-

ing in English. In Proceedings of the First International Conference on Speech Prosody,Bernard Bel and Isabelle Marlien (eds.), 643–646. Aix-en-Provence: LPL – CNRS / SproSig.

— (2003). Sentence phonology. In Oxford Encyclopedia of Linguistics, 2nd ed., Oxford: OxfordUniversity Press.

— (2005). Comments on intonational phrasing. In Prosodies (With Special Reference to IberianLanguages), Sónia Frota, Marina Vigário and Maria João Freitas (eds.), 11–58. Berlin: Mou-ton de Gruyter.

— (2006a). Contrastive focus, givenness and phrase stress. Ms., University of MassachusettsAmherst.

— (2006b). Contrastive focus, givenness and the unmarked status of “discourse-new”. In Inter-disciplinary Studies on Information Structure (ISIS): Working Papers of SFB632, Ausgabe 6,Caroline Féry and Gisebert Fanselow, (eds.), 124–146. Universität Potsdam.

— (2006c). Strong minimalist spellout of prosodic phrases. Paper presented at the Workshop onProsodic Phrasing, GLOW 2006, Universitat Autònoma de Barcelona.

Selkirk, Elisabeth and Angelika Kratzer (2005). Focuses, phases and phrase stress. Paper presentedat the Mediterranean Syntax Meeting, Rhodes.

Selkirk, Elisabeth and Tong Shen (1990). Prosodic domains in Shanghai Chinese. In The Phon-ology–Syntax Connection, Sharon Inkelas and Draga Zec (eds.), 313–338. Chicago: Univer-sity of Chicago Press.

Tokizaki, Hisao (1999). Prosodic phrasing and bare phrase structure. Proceedings of the NorthEast Linguistics Society 29, Pius Tamanji, Masako Hirotani, and Nancy Hall (eds.), Vol. 1,381–395. GLSA, University of Massachusetts Amherst.

— (2006). Linearizing structure with silence: A minimalist theory of syntax–phonology inter-face. Unpublished Ph.d. dissertation, University of Tsukuba.

Truckenbrodt, Hubert (1995). Phonological phrases: their relation to syntax, focus and prominence.Unpublished Ph.D. dissertation, MIT.

— (1999). On the relation between syntactic phrases and phonological phrases. Linguistic In-quiry 30 (2): 219–255.

— (2005). Phrasal stress. In Encyclopedia of Language and Linguistics, 2nd ed., Keith Brown(ed.), 9: 572–582. Oxford: Elsevier.

— (2006). Phrasal prosodic structure in German in cross-linguistic perspective. Paper presentedat the Prosody–Syntax Interface Workshop, University College London.

Uhmann, Susanne (1991). Fokusphonologie: Eine Analyse deutscher Intonationskonturen im Rah-men der nicht-linearen Phonologie. Tübingen: Niemeyer.

Wagner, Michael (2005). Prosody and recursion. Unpublished Ph.D. dissertation, MIT.— (2006). Givenness and locality. In Proceedings of Semantics and Linguistic Theory XVI,

Masayuki Gibson and Jonathan Howell (eds.). Ithaca: CLC Publications (http://research.nii.ac.jp/salt16/).

Wurmbrand, Susi (2005). Verb clusters, verb raising, and restructuring. In The Blackwell Compan-ion to Syntax, Martin Everaert and Hen van Riemsdijk (eds.), 227–341. Oxford: BlackwellPublishers.

Yip, Moira (2002). Tone. Cambridge: Cambridge University Press.

P h a s e th eor y an d p ro s odi c s pe llou t: T h e c...

Documents

Transcript of P h a s e th eor y an d p ro s odi c s pe llou t: T h e c...